- Email: [email protected]
Perinatal intervention trial in Africa: effect of a birth canal cleansing intervention to prevent HIV transmission
Perinatal intervention trial in Africa: effect of a birth canal cleansing intervention to prevent HIV transmission Paolo G Miotti, Taha E Taha, Laban Mtimavalye, Robin Broadhead, Aafke Justesen, Frances Yellin, Liomba, Wendel Miley, David Waters, John D Chiphangwi, James J Goedert
Robert J
George
Biggar,
Introduction
Summary Perinatal transmission of human Background immunodeficiency virus (HIV) type 1 contributes significantly to infant mortality. Exposure in the birth canal may account for some transmission. We examined the efficacy of a birth canal washing procedure in reducing perinatal transmission in Malawi.
Methods The infection status of infants of 3327 control women
(conventional delivery procedures)
was
compared
with that of 3637 infants of intervention-delivered women. The infants’ HIV status was determined by polymerase chain reaction on dried blood spots collected at 6 and 12 weeks of age. The intervention consisted of manual cleansing of the birth canal with a cotton pad soaked in 0·25% chlorhexidine, which was done on admission in labour and every 4 h until delivery.
Findings No adverse reactions to the intervention procedure were seen. 2094 (30%) of the enrolled women HIV-infected, and 59% of their infants were seen in follow-up. Among 982 vaginal vertex singleton deliveries to
were
HIV-infected women, 269 (27%) infants were infected. The intervention had no significant impact on HIV transmission rates (27% in 505 intervention women compared with 28% in 477 control women), except when membranes were ruptured more than 4 h before delivery (transmission 25% in the intervention group vs 39% in the control group). If birth canal exposure is an important risk or additional methods to reduce the risk of perinatal HIV transmission should be tested. Alternatively, perhaps birth canal exposure is not a major contributor to perinatal infection risk.
Interpretation
factor, different
of Africa, the prevalence of human immunodeficiency virus infection in women of childbearing age exceeds 30%,’- 15-45% of infants born to infected women have been reported to become HIVinfected. 1,4 Safe, low-cost, and simple methods should be developed to protect these infants from infections; these methods depend greatly on the timing and routes of exposure.5,6 Zidovudine given to the mother during pregnancy, and to the infant postnatally, can be effective in reducing vertical transmission of HIV .7However, pharmacological strategies are impractical in most of the world because they are costly, require determination of the HIV status of pregnant women and, at least as now used, need to be given for a long period. In the majority of cord blood samples, HIV is not detected; this has been established even in instances where the infant is subsequently found to be infected.9,10 Many investigators therefore believe that transmission occurs most often at delivery, although some transmission has been shown to occur in utero",12 and through breastfeeding. 13,14 Studies of twins indicated that the firstborn infants had at least a two-fold risk of infection in comparison with second-born infants, and in comparison with second-born caesarean-delivered infants, the risk for vaginally delivered first-borns was four-fold higher. 15,16 This observation suggested that birth canal exposure might be an important route of infection. Given this evidence, we postulated that antiseptic cleansing of the birth canal might reduce the risk of perinatal infection much as delivery of the first twin appears to reduce risk for the second twin. We therefore undertook a clinical trial of the safety and efficacy of canal washing as a means to reduce HIV transmission. The study was conducted at Queen Elizabeth Central Hospital, Blantyre, Malawi, where approximately 30% of women giving birth were known to be HIV-infected. 17 In
some
urban
areas
Patients and methods pilot study (160 women) provided evidence that the procedure was safe and that it reduced the vaginal bacterial flora.18 A final protocol was then approved by committees in Malawi and the USA. We expected to enrol about 7000 women and anticipated that this large number of subjects would create record-linkage and misclassification problems if individual randomisation was attempted. Therefore, we blocked enrolment by time period: 2 months (June and July, 1994) in which all delivering women would be delivered conventionally; 3 months (August until October, 1994) in which delivering women were intended to receive the intervention; and a final month (November, 1994) in which the cleansing solutions were withdrawn from the delivery wards and conventional delivery procedures were used. Hereafter, those giving birth in June, July, and November are considered as the control group and those giving birth in August to October are considered as the intervention group. During both control and intervention periods, data about the delivering mothers-their pregnancy, labour and delivery factors, and the A
Viral
Epidemiology Branch, National Cancer Institute, Bethesda, Maryland, USA (R J Biggar MD, J J Goedert MD); Department of Epidemiology, Johns Hopkins School of Public Health, Baltimore (P G Miotti MD); Johns Hopkins University-Ministry of Health Research Project, Blantyre, Malawi (T E Taha MB); Department of Obstetrics and Gynaecology, Queen Elizabeth Central Hospital, Blantyre (L Mtimavalye MB, A Justesen MD, J D Chiphangwi MB); Department of Pediatrics, Queen Elizabeth Central Hospital (R Broadhead MB); Computer Sciences Corporation, Rockville, Maryland, USA (F Yellin MD); Department of Pathology, Queen Elizabeth Central Hospital (G Liomba MB); and SIAC, Frederick Cancer Research Center, Frederick, Maryland (W Miley
MS, D Waters
PhD)
Correspondence to: Dr Robert J Biggar, EPN-434, 6130 Executive Blvd, Rockville, MD 20852, USA
infants born-were collected in a similar manner. Data from the 3 control months were combined because their results were similar (data not presented).
1647
Enrolment required obtaining informed consent. Review of prior records indicated that relatively few women had received antenatal care at the hospital and, for those who had, linking of
positive, since a later positive might have resulted from infection via breastfeeding. The frequency of infant deaths within the first weeks of life
antenatal records to their delivery admissions would have been difficult. Because the mothers were in labour when first seen, consent for birth canal washing was sought during labour at a time when the woman was judged best able to concentrate on the interview. At discharge, each woman was re-interviewed in private, the study further described, and permission for both HIV testing and further follow-up sought. For controls, cord blood was routinely obtained but consent for enrolment and HIV testing was sought at the time of discharge, when they were interviewed in private. Cord blood samples from women who did not consent were discarded without being tested. For the intervention group, treatment consisted of a manual cleansing of the vaginal introitus, the length and vault of the birth canal, the cervical os, and the presenting part of the infant with 0-25% chlorhexidine gluconate in sterile water (ie, 0-0028 mol/L final solution), as well as washing of the infant (excluding eyes and ears). Cotton soaked in the solution and sufficiently damp to leave a residual solution in the vagina was wrapped around the examining fingers of a gloved hand. Chlorhexidine was chosen on the basis of an exceptionally good safety record (and some efficacy against group B streptotoccus),19 and because it can neutralise HIV.20 If the woman was judged to be at term and in true labour, birth canal cleansing (including cleansing of the external genitalia) was done before all examinations. The standard monitoring of delivering women included initial examination and re-examination every 4 h. However, a high proportion of women had had previous deliveries and labour often proceeded quickly; more than 70% had rupture of membranes before arrival at the hospital and more than half delivered within 4 h of rupture. Thus, most mothers had only one or two washes. Some women giving birth during the intervention period were not washed because they did not meet eligibility criteria. Women were excluded if they were in preterm labour without membrane rupture, known to have obstetric problems (including those undergoing elective caesarean sections), admitted in second stage labour, or unable to understand the study fully. We also excluded women whose infants died before discharge and those who refused HIV testing. The mother’s HIV status was established by antibody testing of cord blood on site, by a licensed commercial assay (Genetic Systems EIA, Seattle, Washington, USA). Identification as HIV-infected required repeated positive enzyme-linked immunosorbent assay (ELISA) reactions. Weak or equivocal reactions were confirmed by immunoblotting. 35 women (0-5%) whose status was still questionable after immunoblotting were excluded from analysis. Infants were seen at 6 (range 4 to <8) weeks and 12 (8 to 16) weeks, the normal time for their routine well-baby visits. Infants’ HIV status was established by polymerase chain reaction (PCR). The testing was performed in the USA by the Roche Amplicor HIV-1 test (Roche Diagnostic Systems, Inc, Branchburg, ND. Heel-stick blood (50 jjuL per test) collected on filter papers was used. PCR has been reliably used to determine in-utero and perinatal infection,21-24 and its application to blood collection on filter paper.25 Briefly, this approach appeared to be highly accurate in detecting infection in these infants. In vitro, we were always able to detect as few as five viral copies per blood spot in spiked samples. Furthermore, 57 PCR-positive infants have now had HIV-1 antibody tests done after age 15 months and all are positive. In examining data from 669 subjects with two samples taken at least 1 month apart, we found that 97-6% of those with positive initial samples (always after age 4 weeks) had positive follow-up samples and that 96-2% with negative initial samples continued to be negative in the follow-up samples. Because the risk of becoming infected via breastfeeding during this short interval was small, we used the results from the first sample available any time in the follow-up period. When more than one sample was tested from the same infant, we considered as infected only those infants whose initial sample was PCR
could not be assessed because these mothers would not have returned to the clinic and tracing was not attempted. In another study at this hospital, 5% of HIV-infected infants died before the age of 4 weeks. Within the first months of life, few infants would have illnesses related to HIV, and death rates should be similar in both the intervention and control group. To assess the possible public health benefit, the data were considered in an intent-to-treat analysis, in which intervention and control groups were compared, irrespective of whether or how many times the wash procedure was done in the intervention group. Mean rates with one standard deviation, and risk ratios (RR) and 95% confidence intervals (CI)26 are provided. The data were also stratified in various subsets to determine if the procedure was beneficial in some but not all circumstances.
1648
Results
During the study 3881 control and 4078 intervention gave birth, with 3327 controls and 3637 interventions actually enrolled into the study. Excluded from enrolment (table 1) were 554 control and 441
women
intervention women, and the most common reason for exclusion was complicated delivery (8-0% in both groups). A further 2-0% of control and intervention infants died before their mothers were discharged, and enrolment was not attempted. Significantly more control than intervention-group women refused to give consent for HIV testing, requested at the time of discharge, perhaps because they were first approached about the study only at this time and were eager to leave. The obstetric nurses found the cleansing procedure easy. No complaints or adverse effects were attributed to the intervention. The control and intervention infants were similar in 1 min Apgar score (mean 8-4 and 8-5, respectively) and 5 min Apgar score (both 9-8), suggesting that there was no adverse outcome for the infants. Of the 6964 enrolled women, 2094 (30-2%) were HIVinfected. For this analysis, follow-up was truncated 12 weeks after the last control birth and 16 weeks after the
among intervention
rupture of membranes (ROM) status was recorded but the time of rupture was ,nissing in nine intervention and seven control subjects. Table 2: Efficacy of intervention overall and in subgroups
deliveries, the transmission rate was similar in women with ===4 h (27-8%) and >4 h (25-0%) rupture duration. Among women with rupture of membranes >4 h, the transmission rate was significantly lower for infants born to intervention women (25-0%) than to control women (39-4%) (RR compared with control women with rupture >4 h=0-6; 95% CI 0-4-0-9 [table 2]). There was no correlation between the number of washes and the rate of transmission. Similarly, among infants born to 137 women who were first washed before rupture of membranes the transmission rate was only slightly lower (24-8%). The time between the last wash and delivery did not influence the transmission rate. Other significant univariate correlations with transmission included symptoms of immunodeficiency in the mother, being a first-born infant, low birthweight, and low Apgar score (data not shown). However, the intervention showed no efficacy in any of these subgroups. Discussion
last intervention birth. Within this
1235 (59-0%) HIV-infected women had brought their infants to followup at least once, compared with 61-5% of infants born to
period,
HIV-negative women. The HIV transmission rate varied by delivery factors the intervention. To avoid the variation in HIV transmission rates seen in multiple births, operable deliveries, or complicated deliveries, the effectiveness of the intervention was assessed only in 982 singleton, vertex, vaginally delivered infants, of whom 505 were in the intervention group and 477 were in the control group. Among these 982 infants, the transmission rate was 27-4%. The results reported below refer only to these infants. Follow-up of infants was similar in control and intervention groups (59-3% and 58-7%, respectively). The two groups were not significantly different in terms of mothers’ age (24-2 [SD 54] years), education (6-5 [3’6] years), parity (2-8 [2-1]), and delivery factors, or in the babies’ gestational age (38-0 [2’1] weeks), and extraneous to
birthweight (2869 [468] g). Anvmg deliveries in the intervention period, 136 (26-9%) of 505 infants were found to be infected, compared with 133 (27-9%) of 477 control deliveries (RR 095; 95% CI 0-8-1-2) (table 2). Of women delivering during the intervention period, 59 were not washed for various reasons, and these had a significantly higher transmission rate (39-0%) than those who were washed (RR 1-5; 95% CI 11-22); the majority (91%) had rupture of membranes just before delivery and probably had precipitous deliveries. After exclusion of women in the control period who would have been ineligible for washing had they been in the wash phase, the transmission rate in controls decreased from 27-9% to 26-4%. This was not significantly different from the 25-3% transmission rate in infants of 446 women who were actually washed. Duration of labour was not statistically correlated with transmission rate, although the transmission rate increased as labour lengthened (p,trend,=0-24). However, as shown in table 2, the frequency of transmission increased significantly among controls with duration of membrane rupture, rising from 24-9% in women with rupture 4 h before delivery to 39-4% of women with h (RR 1-6, 95% CI 1-2-2-2). In contrast, rupture >4
This study assessed the efficacy of a simple intervention intended to prevent perinatal HIV transmission. Overall, the 27-4% mother-to-infant transmission rate was similar to that reported in other studies in Africa and elsewhere. 1,3,6 The correlates of HIV transmission were also similar to those seen elsewhere, including higher rates with symptoms of immunodeficiency, low birthweight and, most importantly, rupture of membranes >4 h before delivery. However, there was little evidence that the intervention was effective in preventing HIV transmission. In one subgroup-babies born to women whose membranes were ruptured for >4 h-the intervention may have lowered the risk of HIV infection. Transmission risk was high in the control group with prolonged rupture of membranes in this as well as in other studies.27,28 The association suggests a possible role for ascending infection, a mechanism thought to contribute to bacterial and viral infections (eg, herpes simplex29) of the newborn infant. If so, it could be argued that cleansing of the birth canal lowered the likelihood of ascending infection. However, washing before rupture of membranes should have lowered the content of infectious material before any exposure, but we found no reduction in risk in infants born to women receiving the first intervention before rupture of membranes. In other subgroup analyses, we found that the few women in the intervention period group who did not receive the intervention had a higher risk of transmission; we attribute this to more complicated (mostly precipitous) deliveries. Among those washed, the risk was not progressively reduced by multiple washes. To find out whether reaccumulated blood and mucus containing HIV might have defeated the possible protective effect of washing, we examined transmission rates relative to time between washing and delivery and found no major impact. In sum, the only setting in which washing reduced transmission rates was among infants born to women who had rupture of membranes >4 h before seen in women with delivery. Although the reduction longer durations of rupture was statistically significant, this finding may have been due to chance, which can when multiple subgroup analyses are performed. Overall, the vaginal cleansing intervention used in this trial resulted in only a marginal reduction in the HIV occur
1649
transmission effective
and health
rate
public
be recommended as an approach to reduce HIV perinatal cannot
transmission. This type of intervention has been suggested to reduce neonatal group B streptococcus infection;’9 our pilot study showed a reduction in vaginal bacterial carriage18 and preliminary analyses suggest a lower rate of maternal and neonatal sepsis. The failure of this intervention to reduce HIV transmission could be due to several factors: < Although all nurses were trained to the protocol and monitored, this was a busy clinical setting (30 to 50 deliveries daily) and there may have been inadequate cleansing. We deliberately sought to conduct this study in such a busy setting so as to assess its impact in actual use. Even when done carefully, the cleansing may not have contacted all virus-infected areas, because the presence of infant parts (or the membranes) blocked access to the upper cervical canal. < The chlorhexidine solution might have been ineffective since the lowest virucidal concentration (0-25%) was used. A stronger solution would have been preferable but our foremost consideration was safety and acceptability of the procedure to the mother. We observed that manual cleansing with this approach removed at least some of the potentially infectious blood and mucus in the birth canal. Other antiseptic solutions (eg, benzalkonium chloride or nonoxynol-9) and procedures (eg, irrigation or suppository) should be considered. * It is possible that we overestimated the proportion of perinatal infections attributable to birth canal exposure to HIV. Perhaps transmission in utero-or through other peripartum routes, such as placental bleeding or by instrumentation-has a larger role than is believed. Difficulties also arise in quantifying early peripartum transmission in infants who are breastfed by HIVinfected mothers. If infants become infected soon after birth through HIV-contaminated colostrum or milk, this mode of infection could not have been distinguished from birth canal exposure. A trial of bottle-feeding will clarify this issue, but bottle-feeding in Africa may result in gastrointestinal infections and cannot be sustained in resource-poor areas. The cleansing procedure alone could not have had a major impact if infection in utero or through peripartum routes not within the birth canal is more common than anticipated, or if very early breastfeeding transmission is a major contributor to infection risk. These multiple modes of mother-to-child transmission may make it difficult to find a single effective, yet affordable, public health intervention for
use
in
developing
countries.
This work
was supported in part by a contract from the National Cancer Institute, Bethesda, Maryland, USA (NOI-CP-33005). The support of the PAVE and HIVNET Programs, Division of AIDS, National Institute of Allergy and Infectious Diseases (NO1-AI)-35173-117) has been critical
providing the infrastructure necessary for this project and supplemental funding for the follow-up of infants. We thank the Ministry of Health of Malawi, the staff of the Johns Hopkins-Ministry of Health Research Project, and the nurses at Queen Elizabeth Central Hospital for their active and dedicated collaboration. Special thanks are due to Len van der Hoeven, and Margaret Mhlanga, for their tireless efforts. to
References 1 2
3
Miotti PG, Chiphangwi JD, Dallabetta GA. The situation in Africa. Baillieres Clin Obst Gynaecol 1992; 6: 165-86. Allen S, Van de Perre P, Serufilira A, et al. HIV and malaria in a representative sample of childbearing women in Kigali, Rwanda. J Infect Dis 1991; 164: 67-71. Working Group on Mother-to-Child Transmission of HIV. Rates of
1650
mother-to-child transmission of HIV-1 in Africa, America and Europe: results from 13 perinatal studies. J Acq Immuno Defic Syn Hum Retrovir 1995; 8: 506-10. 4 European Collaborative Study. Children born to women with HIV-1 infection: natural history and risk of transmission. Lancet 1991; 337: 253-60. 5 Borkowsky W, Krasinski K. Perinatal human immunodeficiency virus infection: ruminations on mechanisms of transmission and methods of intervention. Pediatrics 1992; 90: 133-36. 6 Report of a consensus workshop, Siena, Italy, June 3-6, 1993. Strategies for prevention of perinatal transmission of HIV; Siena Consensus Workshop II. J Acq Immuno Defic Syn Hum Retrovir 1995; 8: 161-75. 7 Connor EM, Sperling RS, Gelber R, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type 1 with zidovudine treatment. Pediatric AIDS Clinical Trial Group Protocol 076 Study Group. N Engl J Med 1994; 331: 1173-80. 8 Boyar PJ, Dillon M, Navaie M, et al. Factors predictive of maternalfetal transmission of HIV-1. Preliminary analysis of zidovudine given during pregnancy and/or delivery. JAMA 1994; 271: 1925-30. 9 Bryson YJ, Luzuriaga K, Sullivan JL, Wara DW. Proposed definition for in utero versus intrapartum transmission of HIV-1. N Engl J Med 1992; 327: 1246-47. 10 Simonon A, Lepage P, Karita E, et al. An assessment of the timing of mother-to-child transmission of human immunodeficiency by polymerase chain reaction. J Acq Imm Defic Syn 1994; 7: 952-57. 11 Courgnaud V, Laure F, Broussard A, et al. Frequent and early in utero HIV-infection. AIDS Res Hum Retroviruses 1991; 7: 337-41. 12 Papiernik M, Brossard Y, Mulliez N, et al. Thymic abnormalities in fetuses aborted from human immunodeficiency virus type 1 seropositive women. Pediatrics 1992; 89: 297-301. 13 Dunn DT, Newell ML, Ades AE, Peckam CS. Risk of human immunodeficiency virus type 1 transmission through breastfeeding. Lancet 1992; 340: 585-88. 14 Hira SK, Kamanga J, Bhat GJ, et al. Perinatal transmission of HIV-1 in Zambia. BMJ 1989; 299: 1250-52. 15 Goedert JJ, Duliege AM, Amos CI, Felton S, Biggar RJ. High risk of HIV-1 infection for first-born twins. Lancet 1991; 338: 1471-75. 16 Duliege AM, Amos CI, Felton S, Biggar RJ, Goedert JJ. Birth order, delivery route and concordance in the transmission of human immunodeficiency virus type 1 from mothers to twins. J Pediatr 1995; 216: 625-32. 17 Taha TE, Canner JK, Wangel A-M, et al. Research in HIV in Malawi: the Johns Hopkins University-Ministry of Health Project. Malawi Med J 1994; 10: 6-10. 18 Taha TE, Justesen A, Paterson K, et al. An intervention to reduce the risk of mother-to-child HIV transmission: results of a pilot "toxicity" study. East Afr Med J 1994; 71: 712-15. 19 Burman LG, Christensen P, Christensen K, et al. Prevention of excess neonatal morbidity associated with group B streptococci by vaginal chlorhexidine disinfection during labour. Lancet 1992; 340: 65-69. 20 Harbison MA, Hammer SM. Inactivation of human immunodeficiency virus by betadine products and chlorhexidine. J Acquir Immune Defic Synd 1989; 2: 16-20. 21 Dunn DT, Brandt CD, Krivine A, et al. The sensitivity of HIV-1 DNA polymerase chain reaction in the neonatal period and the relative contributions of intra-uterine and intra-partum transmission. AIDS 1995; 9: F7-F11. 22 Edwards JR, Ulrich PP, Weintrub PS, et al. Polymerase chain reaction compared with concurrent viral cultures for rapid identification of human immunodeficiency virus infection among high-risk infants. J Pediatr 1989; 115: 200-03. 23 Rogers MF, Ou CY, Rayfield M, et al. Use of the polymerase chain reaction for the detection of the proviral sequences of human immunodeficiency virus in infants born to seropositive mothers. N Engl
J Med 1989; 320: 1649-54. 24 Scarlatti G, Lombari V, Plebani A, et al. Polymerase chain reaction, virus isolation and antigen assay in HIV-1 antibody positive mothers and their children. AIDS 1991; 5: 1173. 25 Yourno J, Conroy J. A novel polymerase chain reaction method for detection of human immunodeficiency virus in dried blood spots on filter paper. J Clin Microbiol 1992; 30: 2887-92. 26 Mehta CR, Patel NR, Gray R. Computing an exact confidence interval for the common odds ratio in several 2× 2 contingency tables. J Am Stat Assoc 1985; 80: 969-73. 27 Lepage P, van de Perre P, Msellati P, et al. Mother-to-child transmission of human immunodeficiency virus (type-1) and its determinates in Kigali, Rwanda. Am J Epidemiol 1993; 137: 589-99. 28 Minkoff H, Burns DN, Landesman S, et al. The relationship of the duration of ruptured membranes to vertical transmission of HIV Am J Obstet Gynecol 1995; 173: 585-89. 29 Amstey MS, Monif GRG. Genital herpesvirus infection in pregnancy. Obstet Gynecol 1974; 44: 394-97.
Robert J
George
Biggar,
Introduction
Summary Perinatal transmission of human Background immunodeficiency virus (HIV) type 1 contributes significantly to infant mortality. Exposure in the birth canal may account for some transmission. We examined the efficacy of a birth canal washing procedure in reducing perinatal transmission in Malawi.
Methods The infection status of infants of 3327 control women
(conventional delivery procedures)
was
compared
with that of 3637 infants of intervention-delivered women. The infants’ HIV status was determined by polymerase chain reaction on dried blood spots collected at 6 and 12 weeks of age. The intervention consisted of manual cleansing of the birth canal with a cotton pad soaked in 0·25% chlorhexidine, which was done on admission in labour and every 4 h until delivery.
Findings No adverse reactions to the intervention procedure were seen. 2094 (30%) of the enrolled women HIV-infected, and 59% of their infants were seen in follow-up. Among 982 vaginal vertex singleton deliveries to
were
HIV-infected women, 269 (27%) infants were infected. The intervention had no significant impact on HIV transmission rates (27% in 505 intervention women compared with 28% in 477 control women), except when membranes were ruptured more than 4 h before delivery (transmission 25% in the intervention group vs 39% in the control group). If birth canal exposure is an important risk or additional methods to reduce the risk of perinatal HIV transmission should be tested. Alternatively, perhaps birth canal exposure is not a major contributor to perinatal infection risk.
Interpretation
factor, different
of Africa, the prevalence of human immunodeficiency virus infection in women of childbearing age exceeds 30%,’- 15-45% of infants born to infected women have been reported to become HIVinfected. 1,4 Safe, low-cost, and simple methods should be developed to protect these infants from infections; these methods depend greatly on the timing and routes of exposure.5,6 Zidovudine given to the mother during pregnancy, and to the infant postnatally, can be effective in reducing vertical transmission of HIV .7However, pharmacological strategies are impractical in most of the world because they are costly, require determination of the HIV status of pregnant women and, at least as now used, need to be given for a long period. In the majority of cord blood samples, HIV is not detected; this has been established even in instances where the infant is subsequently found to be infected.9,10 Many investigators therefore believe that transmission occurs most often at delivery, although some transmission has been shown to occur in utero",12 and through breastfeeding. 13,14 Studies of twins indicated that the firstborn infants had at least a two-fold risk of infection in comparison with second-born infants, and in comparison with second-born caesarean-delivered infants, the risk for vaginally delivered first-borns was four-fold higher. 15,16 This observation suggested that birth canal exposure might be an important route of infection. Given this evidence, we postulated that antiseptic cleansing of the birth canal might reduce the risk of perinatal infection much as delivery of the first twin appears to reduce risk for the second twin. We therefore undertook a clinical trial of the safety and efficacy of canal washing as a means to reduce HIV transmission. The study was conducted at Queen Elizabeth Central Hospital, Blantyre, Malawi, where approximately 30% of women giving birth were known to be HIV-infected. 17 In
some
urban
areas
Patients and methods pilot study (160 women) provided evidence that the procedure was safe and that it reduced the vaginal bacterial flora.18 A final protocol was then approved by committees in Malawi and the USA. We expected to enrol about 7000 women and anticipated that this large number of subjects would create record-linkage and misclassification problems if individual randomisation was attempted. Therefore, we blocked enrolment by time period: 2 months (June and July, 1994) in which all delivering women would be delivered conventionally; 3 months (August until October, 1994) in which delivering women were intended to receive the intervention; and a final month (November, 1994) in which the cleansing solutions were withdrawn from the delivery wards and conventional delivery procedures were used. Hereafter, those giving birth in June, July, and November are considered as the control group and those giving birth in August to October are considered as the intervention group. During both control and intervention periods, data about the delivering mothers-their pregnancy, labour and delivery factors, and the A
Viral
Epidemiology Branch, National Cancer Institute, Bethesda, Maryland, USA (R J Biggar MD, J J Goedert MD); Department of Epidemiology, Johns Hopkins School of Public Health, Baltimore (P G Miotti MD); Johns Hopkins University-Ministry of Health Research Project, Blantyre, Malawi (T E Taha MB); Department of Obstetrics and Gynaecology, Queen Elizabeth Central Hospital, Blantyre (L Mtimavalye MB, A Justesen MD, J D Chiphangwi MB); Department of Pediatrics, Queen Elizabeth Central Hospital (R Broadhead MB); Computer Sciences Corporation, Rockville, Maryland, USA (F Yellin MD); Department of Pathology, Queen Elizabeth Central Hospital (G Liomba MB); and SIAC, Frederick Cancer Research Center, Frederick, Maryland (W Miley
MS, D Waters
PhD)
Correspondence to: Dr Robert J Biggar, EPN-434, 6130 Executive Blvd, Rockville, MD 20852, USA
infants born-were collected in a similar manner. Data from the 3 control months were combined because their results were similar (data not presented).
1647
Enrolment required obtaining informed consent. Review of prior records indicated that relatively few women had received antenatal care at the hospital and, for those who had, linking of
positive, since a later positive might have resulted from infection via breastfeeding. The frequency of infant deaths within the first weeks of life
antenatal records to their delivery admissions would have been difficult. Because the mothers were in labour when first seen, consent for birth canal washing was sought during labour at a time when the woman was judged best able to concentrate on the interview. At discharge, each woman was re-interviewed in private, the study further described, and permission for both HIV testing and further follow-up sought. For controls, cord blood was routinely obtained but consent for enrolment and HIV testing was sought at the time of discharge, when they were interviewed in private. Cord blood samples from women who did not consent were discarded without being tested. For the intervention group, treatment consisted of a manual cleansing of the vaginal introitus, the length and vault of the birth canal, the cervical os, and the presenting part of the infant with 0-25% chlorhexidine gluconate in sterile water (ie, 0-0028 mol/L final solution), as well as washing of the infant (excluding eyes and ears). Cotton soaked in the solution and sufficiently damp to leave a residual solution in the vagina was wrapped around the examining fingers of a gloved hand. Chlorhexidine was chosen on the basis of an exceptionally good safety record (and some efficacy against group B streptotoccus),19 and because it can neutralise HIV.20 If the woman was judged to be at term and in true labour, birth canal cleansing (including cleansing of the external genitalia) was done before all examinations. The standard monitoring of delivering women included initial examination and re-examination every 4 h. However, a high proportion of women had had previous deliveries and labour often proceeded quickly; more than 70% had rupture of membranes before arrival at the hospital and more than half delivered within 4 h of rupture. Thus, most mothers had only one or two washes. Some women giving birth during the intervention period were not washed because they did not meet eligibility criteria. Women were excluded if they were in preterm labour without membrane rupture, known to have obstetric problems (including those undergoing elective caesarean sections), admitted in second stage labour, or unable to understand the study fully. We also excluded women whose infants died before discharge and those who refused HIV testing. The mother’s HIV status was established by antibody testing of cord blood on site, by a licensed commercial assay (Genetic Systems EIA, Seattle, Washington, USA). Identification as HIV-infected required repeated positive enzyme-linked immunosorbent assay (ELISA) reactions. Weak or equivocal reactions were confirmed by immunoblotting. 35 women (0-5%) whose status was still questionable after immunoblotting were excluded from analysis. Infants were seen at 6 (range 4 to <8) weeks and 12 (8 to 16) weeks, the normal time for their routine well-baby visits. Infants’ HIV status was established by polymerase chain reaction (PCR). The testing was performed in the USA by the Roche Amplicor HIV-1 test (Roche Diagnostic Systems, Inc, Branchburg, ND. Heel-stick blood (50 jjuL per test) collected on filter papers was used. PCR has been reliably used to determine in-utero and perinatal infection,21-24 and its application to blood collection on filter paper.25 Briefly, this approach appeared to be highly accurate in detecting infection in these infants. In vitro, we were always able to detect as few as five viral copies per blood spot in spiked samples. Furthermore, 57 PCR-positive infants have now had HIV-1 antibody tests done after age 15 months and all are positive. In examining data from 669 subjects with two samples taken at least 1 month apart, we found that 97-6% of those with positive initial samples (always after age 4 weeks) had positive follow-up samples and that 96-2% with negative initial samples continued to be negative in the follow-up samples. Because the risk of becoming infected via breastfeeding during this short interval was small, we used the results from the first sample available any time in the follow-up period. When more than one sample was tested from the same infant, we considered as infected only those infants whose initial sample was PCR
could not be assessed because these mothers would not have returned to the clinic and tracing was not attempted. In another study at this hospital, 5% of HIV-infected infants died before the age of 4 weeks. Within the first months of life, few infants would have illnesses related to HIV, and death rates should be similar in both the intervention and control group. To assess the possible public health benefit, the data were considered in an intent-to-treat analysis, in which intervention and control groups were compared, irrespective of whether or how many times the wash procedure was done in the intervention group. Mean rates with one standard deviation, and risk ratios (RR) and 95% confidence intervals (CI)26 are provided. The data were also stratified in various subsets to determine if the procedure was beneficial in some but not all circumstances.
1648
Results
During the study 3881 control and 4078 intervention gave birth, with 3327 controls and 3637 interventions actually enrolled into the study. Excluded from enrolment (table 1) were 554 control and 441
women
intervention women, and the most common reason for exclusion was complicated delivery (8-0% in both groups). A further 2-0% of control and intervention infants died before their mothers were discharged, and enrolment was not attempted. Significantly more control than intervention-group women refused to give consent for HIV testing, requested at the time of discharge, perhaps because they were first approached about the study only at this time and were eager to leave. The obstetric nurses found the cleansing procedure easy. No complaints or adverse effects were attributed to the intervention. The control and intervention infants were similar in 1 min Apgar score (mean 8-4 and 8-5, respectively) and 5 min Apgar score (both 9-8), suggesting that there was no adverse outcome for the infants. Of the 6964 enrolled women, 2094 (30-2%) were HIVinfected. For this analysis, follow-up was truncated 12 weeks after the last control birth and 16 weeks after the
among intervention
rupture of membranes (ROM) status was recorded but the time of rupture was ,nissing in nine intervention and seven control subjects. Table 2: Efficacy of intervention overall and in subgroups
deliveries, the transmission rate was similar in women with ===4 h (27-8%) and >4 h (25-0%) rupture duration. Among women with rupture of membranes >4 h, the transmission rate was significantly lower for infants born to intervention women (25-0%) than to control women (39-4%) (RR compared with control women with rupture >4 h=0-6; 95% CI 0-4-0-9 [table 2]). There was no correlation between the number of washes and the rate of transmission. Similarly, among infants born to 137 women who were first washed before rupture of membranes the transmission rate was only slightly lower (24-8%). The time between the last wash and delivery did not influence the transmission rate. Other significant univariate correlations with transmission included symptoms of immunodeficiency in the mother, being a first-born infant, low birthweight, and low Apgar score (data not shown). However, the intervention showed no efficacy in any of these subgroups. Discussion
last intervention birth. Within this
1235 (59-0%) HIV-infected women had brought their infants to followup at least once, compared with 61-5% of infants born to
period,
HIV-negative women. The HIV transmission rate varied by delivery factors the intervention. To avoid the variation in HIV transmission rates seen in multiple births, operable deliveries, or complicated deliveries, the effectiveness of the intervention was assessed only in 982 singleton, vertex, vaginally delivered infants, of whom 505 were in the intervention group and 477 were in the control group. Among these 982 infants, the transmission rate was 27-4%. The results reported below refer only to these infants. Follow-up of infants was similar in control and intervention groups (59-3% and 58-7%, respectively). The two groups were not significantly different in terms of mothers’ age (24-2 [SD 54] years), education (6-5 [3’6] years), parity (2-8 [2-1]), and delivery factors, or in the babies’ gestational age (38-0 [2’1] weeks), and extraneous to
birthweight (2869 [468] g). Anvmg deliveries in the intervention period, 136 (26-9%) of 505 infants were found to be infected, compared with 133 (27-9%) of 477 control deliveries (RR 095; 95% CI 0-8-1-2) (table 2). Of women delivering during the intervention period, 59 were not washed for various reasons, and these had a significantly higher transmission rate (39-0%) than those who were washed (RR 1-5; 95% CI 11-22); the majority (91%) had rupture of membranes just before delivery and probably had precipitous deliveries. After exclusion of women in the control period who would have been ineligible for washing had they been in the wash phase, the transmission rate in controls decreased from 27-9% to 26-4%. This was not significantly different from the 25-3% transmission rate in infants of 446 women who were actually washed. Duration of labour was not statistically correlated with transmission rate, although the transmission rate increased as labour lengthened (p,trend,=0-24). However, as shown in table 2, the frequency of transmission increased significantly among controls with duration of membrane rupture, rising from 24-9% in women with rupture 4 h before delivery to 39-4% of women with h (RR 1-6, 95% CI 1-2-2-2). In contrast, rupture >4
This study assessed the efficacy of a simple intervention intended to prevent perinatal HIV transmission. Overall, the 27-4% mother-to-infant transmission rate was similar to that reported in other studies in Africa and elsewhere. 1,3,6 The correlates of HIV transmission were also similar to those seen elsewhere, including higher rates with symptoms of immunodeficiency, low birthweight and, most importantly, rupture of membranes >4 h before delivery. However, there was little evidence that the intervention was effective in preventing HIV transmission. In one subgroup-babies born to women whose membranes were ruptured for >4 h-the intervention may have lowered the risk of HIV infection. Transmission risk was high in the control group with prolonged rupture of membranes in this as well as in other studies.27,28 The association suggests a possible role for ascending infection, a mechanism thought to contribute to bacterial and viral infections (eg, herpes simplex29) of the newborn infant. If so, it could be argued that cleansing of the birth canal lowered the likelihood of ascending infection. However, washing before rupture of membranes should have lowered the content of infectious material before any exposure, but we found no reduction in risk in infants born to women receiving the first intervention before rupture of membranes. In other subgroup analyses, we found that the few women in the intervention period group who did not receive the intervention had a higher risk of transmission; we attribute this to more complicated (mostly precipitous) deliveries. Among those washed, the risk was not progressively reduced by multiple washes. To find out whether reaccumulated blood and mucus containing HIV might have defeated the possible protective effect of washing, we examined transmission rates relative to time between washing and delivery and found no major impact. In sum, the only setting in which washing reduced transmission rates was among infants born to women who had rupture of membranes >4 h before seen in women with delivery. Although the reduction longer durations of rupture was statistically significant, this finding may have been due to chance, which can when multiple subgroup analyses are performed. Overall, the vaginal cleansing intervention used in this trial resulted in only a marginal reduction in the HIV occur
1649
transmission effective
and health
rate
public
be recommended as an approach to reduce HIV perinatal cannot
transmission. This type of intervention has been suggested to reduce neonatal group B streptococcus infection;’9 our pilot study showed a reduction in vaginal bacterial carriage18 and preliminary analyses suggest a lower rate of maternal and neonatal sepsis. The failure of this intervention to reduce HIV transmission could be due to several factors: < Although all nurses were trained to the protocol and monitored, this was a busy clinical setting (30 to 50 deliveries daily) and there may have been inadequate cleansing. We deliberately sought to conduct this study in such a busy setting so as to assess its impact in actual use. Even when done carefully, the cleansing may not have contacted all virus-infected areas, because the presence of infant parts (or the membranes) blocked access to the upper cervical canal. < The chlorhexidine solution might have been ineffective since the lowest virucidal concentration (0-25%) was used. A stronger solution would have been preferable but our foremost consideration was safety and acceptability of the procedure to the mother. We observed that manual cleansing with this approach removed at least some of the potentially infectious blood and mucus in the birth canal. Other antiseptic solutions (eg, benzalkonium chloride or nonoxynol-9) and procedures (eg, irrigation or suppository) should be considered. * It is possible that we overestimated the proportion of perinatal infections attributable to birth canal exposure to HIV. Perhaps transmission in utero-or through other peripartum routes, such as placental bleeding or by instrumentation-has a larger role than is believed. Difficulties also arise in quantifying early peripartum transmission in infants who are breastfed by HIVinfected mothers. If infants become infected soon after birth through HIV-contaminated colostrum or milk, this mode of infection could not have been distinguished from birth canal exposure. A trial of bottle-feeding will clarify this issue, but bottle-feeding in Africa may result in gastrointestinal infections and cannot be sustained in resource-poor areas. The cleansing procedure alone could not have had a major impact if infection in utero or through peripartum routes not within the birth canal is more common than anticipated, or if very early breastfeeding transmission is a major contributor to infection risk. These multiple modes of mother-to-child transmission may make it difficult to find a single effective, yet affordable, public health intervention for
use
in
developing
countries.
This work
was supported in part by a contract from the National Cancer Institute, Bethesda, Maryland, USA (NOI-CP-33005). The support of the PAVE and HIVNET Programs, Division of AIDS, National Institute of Allergy and Infectious Diseases (NO1-AI)-35173-117) has been critical
providing the infrastructure necessary for this project and supplemental funding for the follow-up of infants. We thank the Ministry of Health of Malawi, the staff of the Johns Hopkins-Ministry of Health Research Project, and the nurses at Queen Elizabeth Central Hospital for their active and dedicated collaboration. Special thanks are due to Len van der Hoeven, and Margaret Mhlanga, for their tireless efforts. to
References 1 2
3
Miotti PG, Chiphangwi JD, Dallabetta GA. The situation in Africa. Baillieres Clin Obst Gynaecol 1992; 6: 165-86. Allen S, Van de Perre P, Serufilira A, et al. HIV and malaria in a representative sample of childbearing women in Kigali, Rwanda. J Infect Dis 1991; 164: 67-71. Working Group on Mother-to-Child Transmission of HIV. Rates of
1650
mother-to-child transmission of HIV-1 in Africa, America and Europe: results from 13 perinatal studies. J Acq Immuno Defic Syn Hum Retrovir 1995; 8: 506-10. 4 European Collaborative Study. Children born to women with HIV-1 infection: natural history and risk of transmission. Lancet 1991; 337: 253-60. 5 Borkowsky W, Krasinski K. Perinatal human immunodeficiency virus infection: ruminations on mechanisms of transmission and methods of intervention. Pediatrics 1992; 90: 133-36. 6 Report of a consensus workshop, Siena, Italy, June 3-6, 1993. Strategies for prevention of perinatal transmission of HIV; Siena Consensus Workshop II. J Acq Immuno Defic Syn Hum Retrovir 1995; 8: 161-75. 7 Connor EM, Sperling RS, Gelber R, et al. Reduction of maternal-infant transmission of human immunodeficiency virus type 1 with zidovudine treatment. Pediatric AIDS Clinical Trial Group Protocol 076 Study Group. N Engl J Med 1994; 331: 1173-80. 8 Boyar PJ, Dillon M, Navaie M, et al. Factors predictive of maternalfetal transmission of HIV-1. Preliminary analysis of zidovudine given during pregnancy and/or delivery. JAMA 1994; 271: 1925-30. 9 Bryson YJ, Luzuriaga K, Sullivan JL, Wara DW. Proposed definition for in utero versus intrapartum transmission of HIV-1. N Engl J Med 1992; 327: 1246-47. 10 Simonon A, Lepage P, Karita E, et al. An assessment of the timing of mother-to-child transmission of human immunodeficiency by polymerase chain reaction. J Acq Imm Defic Syn 1994; 7: 952-57. 11 Courgnaud V, Laure F, Broussard A, et al. Frequent and early in utero HIV-infection. AIDS Res Hum Retroviruses 1991; 7: 337-41. 12 Papiernik M, Brossard Y, Mulliez N, et al. Thymic abnormalities in fetuses aborted from human immunodeficiency virus type 1 seropositive women. Pediatrics 1992; 89: 297-301. 13 Dunn DT, Newell ML, Ades AE, Peckam CS. Risk of human immunodeficiency virus type 1 transmission through breastfeeding. Lancet 1992; 340: 585-88. 14 Hira SK, Kamanga J, Bhat GJ, et al. Perinatal transmission of HIV-1 in Zambia. BMJ 1989; 299: 1250-52. 15 Goedert JJ, Duliege AM, Amos CI, Felton S, Biggar RJ. High risk of HIV-1 infection for first-born twins. Lancet 1991; 338: 1471-75. 16 Duliege AM, Amos CI, Felton S, Biggar RJ, Goedert JJ. Birth order, delivery route and concordance in the transmission of human immunodeficiency virus type 1 from mothers to twins. J Pediatr 1995; 216: 625-32. 17 Taha TE, Canner JK, Wangel A-M, et al. Research in HIV in Malawi: the Johns Hopkins University-Ministry of Health Project. Malawi Med J 1994; 10: 6-10. 18 Taha TE, Justesen A, Paterson K, et al. An intervention to reduce the risk of mother-to-child HIV transmission: results of a pilot "toxicity" study. East Afr Med J 1994; 71: 712-15. 19 Burman LG, Christensen P, Christensen K, et al. Prevention of excess neonatal morbidity associated with group B streptococci by vaginal chlorhexidine disinfection during labour. Lancet 1992; 340: 65-69. 20 Harbison MA, Hammer SM. Inactivation of human immunodeficiency virus by betadine products and chlorhexidine. J Acquir Immune Defic Synd 1989; 2: 16-20. 21 Dunn DT, Brandt CD, Krivine A, et al. The sensitivity of HIV-1 DNA polymerase chain reaction in the neonatal period and the relative contributions of intra-uterine and intra-partum transmission. AIDS 1995; 9: F7-F11. 22 Edwards JR, Ulrich PP, Weintrub PS, et al. Polymerase chain reaction compared with concurrent viral cultures for rapid identification of human immunodeficiency virus infection among high-risk infants. J Pediatr 1989; 115: 200-03. 23 Rogers MF, Ou CY, Rayfield M, et al. Use of the polymerase chain reaction for the detection of the proviral sequences of human immunodeficiency virus in infants born to seropositive mothers. N Engl
J Med 1989; 320: 1649-54. 24 Scarlatti G, Lombari V, Plebani A, et al. Polymerase chain reaction, virus isolation and antigen assay in HIV-1 antibody positive mothers and their children. AIDS 1991; 5: 1173. 25 Yourno J, Conroy J. A novel polymerase chain reaction method for detection of human immunodeficiency virus in dried blood spots on filter paper. J Clin Microbiol 1992; 30: 2887-92. 26 Mehta CR, Patel NR, Gray R. Computing an exact confidence interval for the common odds ratio in several 2× 2 contingency tables. J Am Stat Assoc 1985; 80: 969-73. 27 Lepage P, van de Perre P, Msellati P, et al. Mother-to-child transmission of human immunodeficiency virus (type-1) and its determinates in Kigali, Rwanda. Am J Epidemiol 1993; 137: 589-99. 28 Minkoff H, Burns DN, Landesman S, et al. The relationship of the duration of ruptured membranes to vertical transmission of HIV Am J Obstet Gynecol 1995; 173: 585-89. 29 Amstey MS, Monif GRG. Genital herpesvirus infection in pregnancy. Obstet Gynecol 1974; 44: 394-97.